1.Mobile AP

1 Mobile Embedded SoC Technologies for Future User Experiences SeogJun Lee, Ph.D. Vice President AP development Samsung Electronics, Co., Ltd. seogjun.lee@samsung.com Samsung Confidential Samsung Confidential seogjun.lee@samsung.com Aug. 27th, 2012 Technologies for High Performance AP Closing Remarks Mobile AP Introduction Device Trends and Challenges Efforts for Power and Thermal Optimization 3Samsung Confidential seogjun.lee@samsung.com Aug. 27th, 2012 Mobile Application Processor (AP)  Mobile AP is Integrated System-on-Chip which  activates OS and applications for Smartphone and Tablet PC (CPU)  controls system peripherals and interfaces (Chip) [Source: boltportland.com/blog/blog-page] PC Main Board Core (CPU) Graphics Accelerator Engine Memory, Display system/Controller Multimedia en/decoding codec ISP (Image Signal Processor) Camera, Audio, Modem High/Low speed Serial/Parallel connectivity interfaces 4Samsung Confidential seogjun.lee@samsung.com Aug. 27th, 2012 AP and Other Devices Application Processor Memories Image Sensors Display Panels NAND 5Samsung Confidential seogjun.lee@samsung.com Aug. 27th, 2012 Key Components of Mobile AP Wide I/O & TSVAdvanced Package POP Operating System Middleware Applications Software Solutions M/M Codec Memory I/F Display I/FStorage I/F CPU GPU ISP Connectivity Camera I/FIP Technology Cell Library / SRAM Design Methodology Custom Circuit Advanced Low Power Process Design Infrastructure Device Drivers 6Samsung Confidential seogjun.lee@samsung.com Aug. 27th, 2012 More User Experience with AP Samsung Confidential seogjun.lee@samsung.com Aug. 27th, 2012 Technologies for High Performance AP Closing Remarks Mobile AP Introduction Device Trends and Challenges Efforts for Power and Thermal Optimization 8Samsung Confidential seogjun.lee@samsung.com Aug. 27th, 2012 High Performance with Mobile CPU Conventional ArchitectureARM CPU Development Implementing Mobile CPU 800MHz ARM 11 CA8 CA9 CA15 1.0GHz 1.5GHz 2.0GHz ~  High performance CPU essential for rapid response and execution  Standard ASIC design methodology is not sufficient  PPA (Performance, Power, Area) optimization essential  Design methodology can be applied to other high performance IP  Advanced low power CMOS Process : 45nm  32nm  28nm  ….  High speed circuit Implementation  Special high-speed cell library  Custom designed SRAMs and register files  Layout technique for achieving high frequency  Semi-custom design with static and dynamic circuits  Experts for CPU implementation CPU Will be covered at “Application Processor Technology” Session 9Samsung Confidential seogjun.lee@samsung.com Aug. 27th, 2012 GPU Introduction GPU GPU : Graphic Processing Unit designed to rapidly process vector format data to accelerate the building of images intended for output to a display  Architecture for algorithms where processing of large blocks of data is done in parallel  Highly parallel structure : more effective than general-purpose CPU for vector processing Gaming Everywhere!! 10Samsung Confidential seogjun.lee@samsung.com Aug. 27th, 2012 Emerging Trends in Industry  Proliferation of high performance GPU in mobile AP  Graphic intensive applications are now expected to run seamlessly in low power mobile devices  General purpose computing on GPUs (GPGPU)  Parallel processing capability of GPU can benefit general purpose processing scenario  Supporting general purpose instructions in GPU for better programmability GPU 11Samsung Confidential seogjun.lee@samsung.com Aug. 27th, 2012 Consumer Devices and GPGPU GPU Image post-processing and enhancement Image editing software Augmented reality Support emerging new coding schemes quickly • Streaming math/algorithm libraries • Physics modeling • Gaming engines and WOW effects 12Samsung Confidential seogjun.lee@samsung.com Aug. 27th, 2012 Communication Trend in Modem Modem  Modem: Modulator + Demodulator EDGE WCDMA HSPA HSPA+ LTE LTE-A Max. Data Rate (DL/UL, bps) 473K 2M/384K 14M/5.7M 28M/11M 300M/75M 1G/500M Bandwidth(MHz) 0.2 5 5 5 1.4/3/5/10/ 15/20 1.4/3/5/10/ 15/20 U-Plane Latency (msec) 150 150 50 25 <10 <10 2G 2G transitional 3G 3G transitional 4G GSM/3GPP family 13Samsung Confidential seogjun.lee@samsung.com Aug. 27th, 2012 Ever Increasing Multimedia Demands Popup notifications Turn by turn navigation Google’s Project Glasses [Source: www.cnet.com / www.foxnews.com] Refocus Multimedia 14Samsung Confidential seogjun.lee@samsung.com Aug. 27th, 2012 3D & UHD Imaging is Increasing Changing the means of communication beyond Voice 3D Capture Theater Home Personal Full-HD/UHD Movie with continued high resolution High Sensitivity Low Noise Pixel Your phone is starting to replace your DSC One button capture needs highly automated processing Cleaner, Faster, Brighter Galaxy-S User Trends in Imaging Multimedia 15Samsung Confidential seogjun.lee@samsung.com Aug. 27th, 2012  Bus is a subsystem that transfers data between components. On-chip Network ?Crossbar ArchitectureShared Bus Architecture Current & Future?Past (’05 ~ )Past (~’04) ? AXI BusAHB SiliconBackplaneTM  Only one single master can use an on-chip bus  Multiple masters can use an on-chip bus at the same time  Better scalability  Higher clock frequency  Wire efficient or routability 16Samsung Confidential seogjun.lee@samsung.com Aug. 27th, 2012 On-Chip Communication Optimization HW Engines (MM/Comm/Gfx) Memories Processor & DSP “On-Chip Interconnect is a Data Highway to connect IPs”  Only smart INTERCONNECT enables the design of complex chips  On-Chip Bus as well as Memory Sub-Systems and Computing Engines take a big impact on overall system performance On-chip Network 17Samsung Confidential seogjun.lee@samsung.com Aug. 27th, 2012 QoS : Quality of Service  QoS solution provides a fair arbitration mechanism to best support service requirement of each IP Interconnect Memory • CPU • Latency sensitive P e rf o rm a n ce Latency A fair resource allocation according to certain service policy P e rf o rm a n ce Latency • GPU • Insensitive to Latency P e rf o rm a n ce Latency • Display IP • Either works or not • Best effort IPs • Doing their best! On-chip Network 18Samsung Confidential seogjun.lee@samsung.com Aug. 27th, 2012 High-Data Bandwidth : NoC Bus  Achieved high operating frequency by eliminating duplicated logic at each switch • Separates computation from communication • Replace complex global wires with a structured interconnect network using packetized data. • Better scalability • Higher clock frequency • More wire efficiency/Easier routing On-chip Network 19Samsung Confidential seogjun.lee@samsung.com Aug. 27th, 2012 HD Contents Hollywood studios require a high level security Mobile Banking The number of transactions through mobile devices exponentially increases by year Threat of malware and root-kits Root-kit is able to remotely control your device once it is installed. Data Protection Some information must not be disclosed even in the case that the mobile device is stolen/lost. Secure Applications on AP  There is no secure asset that can be compromised Security 20Samsung Confidential seogjun.lee@samsung.com Aug. 27th, 2012 The challenge of implementing AP security Open Platform Everyone is able to access source codes without any permission. Once a hacker succeeds rooting, he/she is able to control all the resources in the mobile device  Every Rose Has Its Thorn.  Android is open platform using open source linux kernel. Payment & Ticketing Security 21Samsung Confidential seogjun.lee@samsung.com Aug. 27th, 2012 AP Security: TrustZone Security Define Secure Hardware Architecture Normal and Secure Domain Implement in SoC Enforcing secure/normal separation physically Combine SoC with Trusted OS A Trusted Execution Environment (TEE) Securing the system with TrustZone Ready Program, ARM 22Samsung Confidential seogjun.lee@samsung.com Aug. 27th, 2012 AP Security: TrustZone Security  Separate OS as Normal and Secure Domains  Security Extensions of ARM and Security-Aware Components for running Secure OS 23Samsung Confidential seogjun.lee@samsung.com Aug. 27th, 2012 AP Security: TrustZone Security  Separate OS as Normal and Secure Domains  Security Extensions of ARM and Security-Aware Components for running Secure OS Securing the system with TrustZone Ready Program, ARM  Trustzone trusted PIN entry path 1 1 Trusted hardware peripheral input 2 2 To trusted software 3 3 To trusted authentication service 24Samsung Confidential seogjun.lee@samsung.com Aug. 27th, 2012 Multiple OS vendors and the variants Frequent OS version updates Optimization of software parallelization for performance scalability Heterogeneous multi-processing Various multimedia/computing standards (e.g., OpenGL, Direct3D, OpenCL, Renderscript, OpenSL, etc.) Emerging cloud computing standard: HTML5 Better security, power efficiency, and reliability OS Portability Multi-processing Support Standard Support Virtualization Support Software Challenges  Software development needs are rapidly increasing Will be presented at “Software Solution for AP” Session 25Samsung Confidential seogjun.lee@samsung.com Aug. 27th, 2012 Scope of SoC Verification  Verification is getting more important with ever increasing SoC design complexity Architecture exploration Functionality check Performance verification Power estimation Post-silicon verification Will be covered at “The Challenges in Design and Verification of SoCs” Session Samsung Confidential seogjun.lee@samsung.com Aug. 27th, 2012 Technologies for High Performance AP Closing Remarks Mobile AP Introduction Device Trends and Challenges Efforts for Power and Thermal Optimization 27Samsung Confidential seogjun.lee@samsung.com Aug. 27th, 2012 Mobile DRAM Trends Memory Quad-Core CPUs Multi-tasking Larger screens Longer Standby Power Performance Minimalism Small Size, Small Z-height Instant ON  Instant Play RESPONSIVE Mobile Computing 3.7” 4.0” Full HD, 3D 28Samsung Confidential seogjun.lee@samsung.com Aug. 27th, 2012 Web Contents Storage Storage Photos Video Contents E-book  Flash capacity and data usage is doubling every year. Will be covered at “Nand Flash Technology (1,2)” Session Contents Sharing 29Samsung Confidential seogjun.lee@samsung.com Aug. 27th, 2012 Interface Revolution Storage • And, requires Interface Revolution! 1 6 Gbps 3 PC USB Now Future Now Future Now Future PHY 6GBps Potential Bottleneck! External Card Mobile Device 1.5Gbps 480Mpbs 416Mpbs 6Gbps 5Gbps 3Gbps SATA1 SATA3 USB2.0 USB3.0 MMC4.3 UFS1.0 Next Gen. UFS 30Samsung Confidential seogjun.lee@samsung.com Aug. 27th, 2012 CIS Your phone is replacing your DSC One button capture needs highly automated processing Cleaner, Faster, Brighter Portable Camera System Change Analog Era Digital Era Digital Convergence Era 31Samsung Confidential seogjun.lee@samsung.com Aug. 27th, 2012 CISCamera Trends Main Camera Trends Second Camera Trends Smartphone use 2nd Camera as input device for UI. New UI based on Face detection and motion detection Smart phone will encroach on DSC market Smart Stay Gaze Sense Zero Shutter Lag → Full 30fps Best Shots Will be covered at “CMOS Image Sensor Tech” Session Samsung Confidential seogjun.lee@samsung.comAug. 27th, 2012 Technologies for High Performance AP Closing Remarks Mobile AP Introduction Device Trends and Challenges Efforts for Power and Thermal Optimization 33Samsung Confidential seogjun.lee@samsung.com Aug. 27th, 2012 seogjun.lee@samsung.com Clock controller Power manager ABB controller CPU V_high Audio V_low Bus V_typical Level shifter Level shifter DVFS controller Clock Gating Power Gating Voltage Island DVFS (Dynamic Voltage Frequency Scaling) ABB (Adaptive Body Biasing) Adaptive Thermal Management Low power SRAM • Write/read assist • Low leakage transistor Low Power Design Methodology  Multiple techniques are used to effectively reduce power consumption Will be covered at “Modern SoC Design Flow and Methodology” Session 34Samsung Confidential seogjun.lee@samsung.com Aug. 27th, 2012 seogjun.lee@samsung.com Power Optimization: Multiple Power Domain Power Management ARM Video LCD Memory 3D MPEG4 Decoding 400MHz 133MHz 133MHz 400MHz OFF MP3 Playing 133MHz OFF OFF 400MHz OFF Mem 3D Video ARM Other LCD Power Domain The power consumption for MP3 playing is much less than that for MPEG4 decoding. 35Samsung Confidential seogjun.lee@samsung.com Aug. 27th, 2012 seogjun.lee@samsung.com PMIC and AP Power Management PMIC: Power Management IC, which supplies power to AP Power Supply (PMIC) CPU G3D Video/LCD I/O Memory AP Power Management Unit 36Samsung Confidential seogjun.lee@samsung.com Aug. 27th, 2012 seogjun.lee@samsung.com Power Optimization: DVFS Power Management  Energy = Power X Time, Power ∝V 2∙f  DVFS (Dynamic Voltage-Frequency Scaling)  IP blocks do not always require full performance.  Small application program runs at very low frequency.  Power dissipation can be reduced through low operating voltage PMIC control signal PMIC Mobile AP VDD 37Samsung Confidential seogjun.lee@samsung.com Aug. 27th, 2012 DVFS; Dynamic Voltage and Frequency Scaling Standby mode Full CPU/GPU operation time By CPU load 100MHz 1.0GHz ON OFF OFF OFF OFF I/O processor or CPU coprocessor; ISR/Task handling without CPU help OFF OFF ON ON OFF ON ON ON ON OFF Multi-core on and off by the task loads Adaptive load balancing Efficient Power Management by SW Power Management 38Samsung Confidential seogjun.lee@samsung.com Aug. 27th, 2012 Power Optimization: Hot Plug-in Power Management  MPCore extends control over power usage by providing both voltage and frequency scaling and turning off unused processors Minimum Vdd Limits Energy Saving Shutdown Powered Down E n e rg y c o n s u m e d Required performance 1 CPU operation 2 CPU operation 3 CPU operation Fully Powered All processors active: DVFS for required performance Reduced concurrency or lower performance required: shut down processors Energy saving from dynamic voltage and frequency scaling Energy saving from dynamic voltage and frequency scaling low power standby 4 CPU operation 39Samsung Confidential seogjun.lee@samsung.com Aug. 27th, 2012 Limited battery size  Low Power solution Increasing functionalities, Increasing display size /resolution, Need for Fast response  High Performance solution Small form factor, direct contact with human skin  Thermal solution AP SOC Challenges Thermal Management 40Samsung Confidential seogjun.lee@samsung.com Aug. 27th, 2012 A high end Smartphone shows a heat spot with the temperature of 46.3C after a casual work load of phone call, map search, web surfing and gaming of 10min each. (Source : etnews.com) Estimated Time and Temperature Effect of Hot Water on Adult skin (Source : “Study on Thermal Injury : II” American Journal of Pathology 1947) Water temp Time for Deg. C Deg. F 1st Degree Burn 2nd Degree burn 43 110 Normal Hot Shower 47 116 35 Minutes 45 Minutes 49 120 3 Minutes 9 Minutes 50 122 1 Minute 5 Minutes 52 126 30 Seconds 90 Seconds 55 131 5 Seconds 25 Seconds 60 140 2 Seconds 5 Seconds 65 149 1 Second 2 Seconds 68 154 Instantaneous 1 Second Thermal Issue : User Point of View Thermal Management  The external temperature (skin temperature) of high end Smartphone now reaches the point that can cause skin damage Samsung Confidential seogjun.lee@samsung.comAug. 27th, 2012 Technologies for High Performance AP Closing Remarks Mobile AP Introduction Device Trends and Challenges Efforts for Power and Thermal Optimization 42Samsung Confidential seogjun.lee@samsung.com Aug. 27th, 2012 seogjun.lee@samsung.com  High Performance  Long battery Life  Competitive cost